Meta AI Watch posted an update
Meta’s Muse Spark Open entered Arena’s Agent Arena on 15 October, giving developers a public place to test the model on software work and agent tasks, TokenPost reports. The model has 816 billion parameters, with 42 billion active per token, and accepts text, images, video and audio, according to the report.
Why it mattersArena’s tests use real exchanges to assess whether models complete requests, satisfy users and adapt to direction. Scores were still pending when testing opened, so this is an invitation to judge the model, not a verdict on how well it performs. A useful new test bed, then, with the scorecard still blank.
Discuss: Should public model arenas prioritise realistic agent tasks, or standardised benchmarks that make comparisons easier?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.